University of Nebraska - LincolnDigitalCommons@University of Nebraska - Lincoln

U.S. Environmental Protection Agency Papers United States Environmental Protection Agency

2014

Esterification pretreatment of free fatty acid inbiodiesel production, from laboratory to industryMing ChaiGreenleaf Biofuels

Qingshi TuUniversity of Cincinnati

Jeffrey Y. YangU.S. Environmental Protection Agency

Mingming LuUniversity of Cincinnati, LUMG@ucmail.uc.edu

Follow this and additional works at: http://digitalcommons.unl.edu/usepapapers

This Article is brought to you for free and open access by the United States Environmental Protection Agency at DigitalCommons@University ofNebraska - Lincoln. It has been accepted for inclusion in U.S. Environmental Protection Agency Papers by an authorized administrator ofDigitalCommons@University of Nebraska - Lincoln.

Chai, Ming; Tu, Qingshi; Yang, Jeffrey Y.; and Lu, Mingming, "Esterification pretreatment of free fatty acid in biodiesel production,from laboratory to industry" (2014). U.S. Environmental Protection Agency Papers. Paper 212.http://digitalcommons.unl.edu/usepapapers/212

Esterification pretreatment of free fatty acid in biodiesel production,from laboratory to industry

Ming Chai a, Qingshi Tu b, Mingming Lu b,⁎, Y. Jeffrey Yang c

a Greenleaf Biofuels, 100 Waterfront St., New Haven, CT 06512, United Statesb School of Energy, Environment, Biological and Medical Engineering, University of Cincinnati, Cincinnati, OH 45221, United Statesc National Risk Management Research Laboratories, U.S. Environmental Protection Agency, Cincinnati, OH 45221, United States

a b s t r a c ta r t i c l e i n f o

Article history:Received 2 January 2013Received in revised form 12 February 2014Accepted 22 March 2014Available online xxxx

Keywords:Acid-catalyzed esterificationBiodieselUsed cooking oilFree fatty acidMethanol to FFA ratioOptimization

In the US, biodiesel producers usually follow the 19.8:1 methanol-to-FFA molar ratio for free fatty acid (FFA) es-terification, as suggested by the National Renewable Energy Laboratory (NREL) without optimization studies. Inthis paper, both laboratory studies and industrial practices of the esterification process were compared, and anoptimization study of a used vegetable oil with 5% FFA was conducted. The optimal conditions of this oil, i.e.,methanol-to-FFA molar ratio of 40:1, and sulfuric acid usage of 10%, fell out of the suggested range of 19.8:1.The activation energy of the esterification reaction is 20.7 kJ/mol at the optimized condition and 45.9 kJ/mol atthe 19.8:1 methanol to FFA ratio. It was found that the 19.8:1 methanol-to-FFA molar ratio worked well onlywithin the FFA range of 15–25% while the suggested 5% sulfuric acid worked well only within the FFA range of15–35%. Outside these ranges, especially at FFA levels less than 15%, optimization study is necessary. Regressionmodels of methanol and acid dosing have been utilized in two industrial scale biodiesel producing facilities andhave successfully reduced the FFA level to less than 0.5%.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Biodiesel is considered as a “direct-pour” alternative fuel to petro-leumdiesel, as it requires almost nomodification tomostmodern dieselengines. Biodiesel can be produced locally and therefore reduces foreignoil dependence. It has been reported that biodiesel combustion can re-sult in less air pollutant emissions, such as carbon monoxide, sulfur di-oxide, particulate matter, hydrocarbons, but with slightly highernitrogen oxides [1]. Since the feedstockof biodiesel ismostly renewable,it significantly reduces carbon dioxide emission during its whole lifecycle [2]. But the reliance on virgin oil as biodiesel feedstock raised sus-tainability concerns, such as the “Food vs. Fuel” debate [3], and con-sumption of resources such as land and water [4]. In addition, itmakes biodiesel less competitive in the fuel market due to the highcost of virgin oil. The cost of the feedstock usually accounts for morethan 80% of the total cost in biodiesel production [5]. Therefore, themar-ket fraction of biodiesel from virgin oil has decreased in recent years inthe U.S. More and more commercial biodiesel producers are capable of

handling multi-feedstocks which include soybean oil, used cooking oil,and animal fats, etc.

Many studies have reported the production of biodiesel from wasteoil feedstock with a few cited here to summarize the advantages andchallenges of using waste oils [6–9]. The waste feedstock can rangefrom used cooking oil from restaurants to animal fats from renderingcompanies. The utilization of waste oil reduces the feedstock cost andincreases the sustainability of biodiesel production by minimizing re-source consumption [9]. Depending on the cooking process and subse-quent storage, the used oils may contain impurities such as water, foodresidues, and a high free fatty acid (FFA) concentration. Themajor tech-nical challenge of making biodiesel from low-quality used oil is the pre-treatment of FFA. FFA is undesirable during the alkali transesterificationprocess due to the formation of soap, yield loss, and increased difficultyin product separation [9,10]. The earlier practice of caustic stripping is toremove the FFA by forming soap with alkali materials. However, thiscan result in biodiesel yield loss, more alkaline usage and a potentiallydelayed phase separation by the excess soap formation fromneutraliza-tion [11]. The acid-catalyzed transesterification can directly convertboth FFA and oil into biodiesel. However, it is not much practiced bythe biodiesel producers due to the longer reaction time and loweryield [12]. Instead, the two-step conversion process: an acid-catalyzedesterification pretreatment to lower the FFA content followed by thetraditional alkali-catalyzed transesterification [7], is widely used inboth industry and laboratory. The acid-catalyzed esterification requiresadditional acid and methanol usage, but the majority of methanol can

Fuel Processing Technology 125 (2014) 106–113

⁎ Corresponding author at: School of Energy, Environment, Biological and MedicalEngineering, PO Box 210012, University of Cincinnati, Cincinnati, OH 45221, UnitedStates. Tel.: +1 513 556 0996; fax: +1 513 556 2599.

E-mail address: LUMG@ucmail.uc.edu (M. Lu).

http://dx.doi.org/10.1016/j.fuproc.2014.03.0250378-3820/© 2014 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Fuel Processing Technology

j ourna l homepage: www.e lsev ie r .com/ locate / fuproc

be reclaimed through a methanol recovery system, which is nowcommonly installed by biodiesel manufacturers. Table 1 compares thechemical costs and yield loss of caustic stripping and acid-catalyzedesterification pretreatment. The esterification process is currently eco-nomically favored.

For the dosage of FFA pretreatment, most industrial biodiesel manu-facturers adopt the dosing regimen recommended by a NationalRenewable Energy Laboratory (NREL) report, which is 2.25 g of metha-nol and 0.05 g of sulfuric acid for every gramof FFA in the oil (equivalentto 19.8:1 of methanol-to-FFA molar ratio and 5% of acid-to-FFA weightpercentage) [11]. Dosage optimization is generally not performed bythe commercial producers, while optimal operating conditions are usu-ally sought for in laboratory studies and are summarized in Table 2[13–31]. The industrial and laboratory values of methanol and acid dos-ages are not in agreement. The discrepancy is especially significant atFFA levels less than 15%. Farage et al. blended a mixture of soybean oiland sunflower oil at 1:1 ratio with oleic acid to create an oil with8.50% FFA and found that the optimal chemical usage for esterificationwas 24:1 methanol-to-FFA molar ratio and 29.4% (weight relative toFFA) acid usage [24]. Hayyan et al. found that the optimal dosages fora sludge palm oil (contains 23.2% FFA) were methanol-to-FFA molarratio of 13:1 and acid amount of 3.23% (weight relative to FFA) [25]. Op-timization of chemical usage is necessary for the oils with various FFAcontents.

Therefore the goal of this paper is to evaluate the optimal chemicalusage differences between industry and laboratory studies. Optimalconditions of used cooking oil with less than 15% FFA were experimen-tally determined. The results were integrated with existing studies tobetter evaluate the efficacy of the NREL regimen. Regression analyseswere performed to better evaluate the correlation between the initialFFA content and the optimal chemical usage. Kinetic parameters werecalculated to better understand the differences of industrial and labora-tory studies.

2. Materials and methods

2.1. Materials

The used cooking oil was collected from restaurants inside theCincinnati Zoo and Botanical Gardens. The esterification reactionused sulfuric acid (HPLC grade, 99.8%, Pharmco-AAPER) as the cata-lyst and methanol (HPLC grade, 99.9%, Pharmco-AAPER) as thereactant.

2.2. Experimental set-up and procedure

The esterification was performed in a one-liter three-neck round-bottom flask (Ace Glass Inc.). One neck was equipped with a thermom-eter to measure the temperature. A water cooled condenser was

connected to another neck on top of the reactor to reduce evaporativeloss ofmethanol. The third neck is used for chemical addition and takingsamples. The reactor was placed in a water bath and heated on ahotplate (Fisher Scientific, 11-100-100SH). Methanol and H2SO4 weremixed before the reaction and oil were heated to the desired tempera-ture before the addition of methanol and H2SO4 mixture. The methanoland oil are immiscible so the agitating speed was kept at 600 revolu-tions per minute (rpm) to ensure efficient mixing, as suggested by aprevious study [13]. For each run, 400 ml of used cooking oil wasadded into theflask and heated to the desired temperature. 5-ml aliquotof sampleswaswithdrawn from the flask for titration at 5, 10, 20, 30, 60,90, and 120 minute intervals after the onset of the reaction. The titrationof FFA followed AOCS Cd 3d-63, which is a standard method for FFA ti-tration in oil [32].

The effects of three essential operating parameters i.e., reaction tem-perature, methanol dose and acid catalyst dose on the FFA conversionrates were investigated. The studied temperatures were 35, 45, 55 and65 °C (approximately the boiling point of methanol) at atmosphericpressure. The quantities of methanol were expressed as methanol-to-FFA molar ratio that varied from 20:1 to 60:1 with 10:1 interval(equal to 3.1, 4.7, 6.3, 7.8, and 9.4 vol.% to oil). Although the esterifica-tion reaction requires one mole of methanol for one mole of FFA, inpractice excessive methanol is often added since this reaction is revers-ible [13]. The quantity of sulfuric acid was expressed as weight percent-age to FFA and ranged from 5% to 15% with 2.5% increment.

Both the acid value (the absolute value) and the FFA conversion rate(a relative scale) can be used to indicate the completion of acid esterifi-cation reaction. The target acid value suggested is less than 2mg KOH/g(roughly 1% FFA) in order to proceed to the alkali-catalyzedtransesterification. In this study, the initial acid value of theused cookingoil was 10± 1mg KOH/g. Hence the FFA conversion rate of 80%was setas a cutoff point to evaluate the effectiveness of the esterification reac-tion, i.e. whether the end product had reached less than 1% FFA andwas suitable for the subsequent alkali-catalyzed transesterification.The FFA conversion rate was calculated by Eq. (1).

FFA Conversion ¼ Initial FFA−Final FFAð ÞInitial FFA

� 100% ð1Þ

Where:

Initial FFA initial acid value (mg KOH/g)Final FFA final acid value (mg KOH/g)

2.3. Analytical methods

Theproperties of the used cooking oilwere analyzedby theASTMandAOCS standard testing procedures as listed in Tables 3.1 and 3.2. The glyc-erides were analyzed by a Hewlett-Packard gas chromatograph (model5890) with a flame ionization detector and an auto sampler (model7673). A Restek Rtx-Biodiesel TG column (10 m ∗ 0.32 mm ∗ 0.1 um)with a 2 m ∗ 0.53 mm guard column was used. The operating conditionsfollowed the ASTM D6584. A small amount of the oil was convertedinto biodiesel via acid-catalyzed esterification and alkali-catalyzedtransesterification for fatty acid compositional analysis (Table 3.1). Thechemical composition of methyl esters was analyzed by a Hewlett-Packard gas chromatography (model 5890) and mass spectrometry(model 5970) (GC–MS) system with an auto sampler (model 7673).The operating conditions were: Restek Rxi-5 ms column (30 m∗ 0.25 mm ∗ 0.25 um), injector temperature at 250 °C and detector tem-perature at 250 °C, flow rate of helium 1 ml/min, split ratio of 5:1, oventemperature starting at 40 °C with holding time of 2 min, increasing to180 °C at 10 °C/min, then to 230 °C at 5 °C/min, and finally to 300 °C at15 °C/min with holding time of 4 min.

Table 1Cost comparison FFA pretreatment methods.

FFA% in oil 2.5% 5% 10% 15%

Caustic stripping Yield loss 2.5% 5% 10% 15%Cost ($/gal) 0.119 0.238 0.476 0.715Cost ($/L) 0.031 0.063 0.126 0.189

Acid esterification Yield loss 1% 1% 1% 1%Cost ($/gal) 0.040 0.080 0.160 0.240Cost ($/L) 0.011 0.021 0.042 0.063

Notes: The costs of chemicals and sale price of biodiesel are all based on current industrialscale values: prices of sulfuric acid, sodiummethylate andmethanol are $5/gal, $4/gal, and$1.6/gal respectively, and $4/gal of sale price for biodiesel is used (due to various federaland local tax credit and incentives program, the sale price of biodiesel could changedramatically). The methanol recovery efficiency is assumed at 80%.

107M. Chai et al. / Fuel Processing Technology 125 (2014) 106–113

3. Results and discussion

The chemical composition of the used cooking oil is listed inTables 3.1 and 3.2. The fatty acid profile resembled canola oil. The5% FFA content was higher than the allowable level for directalkali-catalyzed transesterification. As used cooking oil, the triglyc-erides level is lower than that of the virgin oil. The FFA is hydroly-sis/oxidation by-products of oil due to cooking and storage, andmonoglyceride and diglycerides are degrade products of oil. Thewater, MIU (moisture, insoluble, and unsaponifiables), phosphorousand sulfur levels of the used cooking oil were all in reasonable rangesfor making biodiesel. For a biodiesel producer, this used oil is of rea-sonably good quality.

3.1. Optimization of the esterification process

Fig. 1 shows the effect of reaction temperature on FFA conversionunder different methanol usage. The reaction time of 2 h was usedbased on industrial practices of the pretreatment time. In industrialpractice, prolonging reaction time can be used to fix a batch but usuallynot favored due to resultant cost increase. All of the four figures sug-gested that the higher the reaction temperature, the more completethe FFA conversion would be. The optimal reaction temperature rangewas 55–65 °C, which is consistent with most studies listed in Table 2,and also consistent with the industrial practice. In practice, biodieselmanufacturers need to balance the reaction time (2 h) and temperature(~60 ºC) to obtain both high yield and low energy consumption.

Table 2Summary of published results on FFA pretreatment.

References Oil type FFA(% weight of oil)

Methanola

(to FFA molar ratio)Acidb

(% weight of FFA)Temperature(°C)

13 Sunflower 2.99 60:1 5 6014 Rubber seed 16.88 13:1 6 45 ± 515 Pongamia pinnata oil 8.13 37:1 12.3 6016 First stepc Mahua oil (crude extracted oil) 19 16:1 10.76e 6016 Second stepc 2.27 103:1 90.06e

17 Jatropha (crude seed oil) 14.9 36:1 6.7 5018 Jatropha 14 24:1 20.88 6019 Palm fatty acid distillate 93.0 8:1 1.83 7020 First stepc Rice bran oil 20 11:1 5e 6020 second stepc 2.4 95:1 42e

21 Crude palm oil 7.5 24:1 (ethanol) 4f Microwave22 Tobacco oil 35 18:1 5.71 6023 Crude palm and rubber seed oil (1:1) 11.90 44:1 4.20 6524 Sunflower and soybean oil (1:1) 8.50 24:1 29.41 6025 Sludge palm oil 23.20 13:1 3.23 6026 Mixed crude palm oil 10 10:1 10 6027 Waste frying oil 1.01 201:1d 67.3 5128 Karanja oil 8.80 34:1 0.42f Microwave29 Sludge palm oil 22.33 17:1 3.36 6030 Waste cooking oil 37.96 18:1 10.54 9531 Waste cooking oil 8.71 23.4:1 5.74f Microwave

In this study, methanol-to-FFAmolar ratio and acid to FFAweight percentage are used instead of reactant to oil ratio or reactant to FFA ratio for easy comparison purposes. The conversionequations are as follows:

Methanol −to− FFAmolar ratio ¼ methanol to oilmolar ratio

MWoil

� �=

FFA%MWFFA

� �

Acid −to− FFA%weight ¼acid to oil%weight

FFA%:

The average molecular weights of FFA and oil are calculated by the acid profiles reported in the literatures. Otherwise, 885.46 g/mol, molecular weight of triolein is used as the averagemolecular weight of oil; and 282.46 g/mol, molecular weight of oleic acid is used as the average molecular weight of FFA [11].

a The ratio from the original publication has been converted into methanol-to-FFA molar ratio.b The acid usage from the original publication has been converted into sulfur acid to FFA wt%.c The literature used 2-step acid-catalyzed esterification to lower the FFA level, the optimal conditions of the first and second esterifications were listed separately.d The optimal methanol usage was determined in another study by the same authors, we don't have the access to the original thesis.e The acid usage was fixed, so the results were not included in the acid usage comparison.f The heating resource was different from all other studies, the results were not included in the acid usage comparison due to the lack of temperature control.

Table 3.1Profile of the fatty acid methyl ester from used cooking oil.

Fatty acid profile Relative wt.%

Lauric acid methyl ester C12:0 0.02%Myristic acid methyl ester C14:0 0.03%Palmitic acid methyl ester C16:0 3.34%Palmitoleic methyl ester acid C16:1 0.13%Stearic acid methyl ester C18:0 2.09%Oleic acid methyl ester C18:1 79.75%Linoleic methyl ester acid C18:2 12.39%Linolenic methyl ester acid C18:3 2.04%Arachidic methyl ester acid C20:0 0.22%Unsaturated methyl esters 94.31%

Table 3.2Chemical analysis of the used cooking oil used in this study.

Test Result Method

Free fatty acid 5.0 wt.% AOCS Cd 3d-63Triglycerides 89.6 ± 1.0 wt.% ASTM D6584Diglycerides 5.2 ± 0.2 wt.%Monoglycerides 1.4 ± 0.2 wt.%Density 0.920 g/ml ASTM D 1298Water 0.23 v/v% AOCS Ca 2e-84Sediment ~0.5 v/v% AOCS Ca 3d-02MIU b1 wt.% AOCS Ca 3a-46 and

AOCS Ca 6b-53Phosphorus 9.0 ppm AOCS Ca 20-99Sulfur 5.6 ppm AOCS Ca 17-01

108 M. Chai et al. / Fuel Processing Technology 125 (2014) 106–113

Meanwhile, Fig. 1 also suggested that FFA reduction is also affectedby the methanol-to-FFA molar ratio. When using 50:1 methanol-to-FFA molar ratio and 10% acid, the 2-hour acid values decreased to lessthan 2 mg KOH/g for all test temperatures. At 40:1 methanol-to-FFAmolar ratio, FFA conversions at 45, 55, and 65 °C met the 2 mg KOH/gtarget, but not at 35 °C (2.2 mg KOH/g). At methanol-to-FFA molarratio of 30:1, the target FFA value under was only met at 55 and 65 °C.Further reducing methanol-to-FFA molar ratio to 20:1, which is theNREL suggested value, the target acid value could not be met at anytemperature.

The impacts of methanol-to-FFA molar ratio on the 2-hour FFAconversion rates are further explained in Fig. 2. The target FFA con-centration of 1% is represented by the 80% FFA conversion rate, andall results shown in this figure were with 10% sulfuric acid concen-tration. The FFA conversion rate increased with methanol-to-FFAmolar ratio increasing from 20:1 to 40:1, regardless of reactiontemperature. Increasing the methanol-to-FFA molar ratio to 50:1 or60:1, the 2-hour FFA conversion rates slightly changed. An ANOVA(analysis of variance) single factor test was performed to evaluateif there were statistically significant differences between FFA con-version rates and methanol-to-FFA molar ratios from 40:1 to 60:1.The p-values at 45 °C, 55 °C, and 65 °C were 0.35, 0.40, and 0.08 re-spectively, all larger than the significance level at 0.05, which indi-cated that changes of conversion rates were not statisticallysignificant when the methanol-to-FFA increased beyond 40:1.Therefore, the optimal methanol-to-FFA molar ratio was determinedas 40:1. This result is qualitatively in agreement with some studies,

which reported that the additional methanol did not improve theconversion beyond a certain ratio [13,14].

Fig. 3 shows the effect of catalyst quantity on FFA conversion rate.The methanol-to-FFA molar ratio was fixed at 40:1 for this plot. Therange of catalyst quantity, 5–15% (weight, relative to FFA), fitted in the

Fig. 1. Effect of reaction temperature on acid value.

Fig. 2. Effect of methanol molar ratio to FFA ratio on the 2-hour FFA conversion rate.

109M. Chai et al. / Fuel Processing Technology 125 (2014) 106–113

ranges reported in Table 2. The conversion rate increased with the cat-alyst usage from 5 to 10%, and then decreased with further increase ofthe acid catalyst to 15%. The optimal acid amount was 10% (weight, rel-ative to FFA). Similar result has also been observed by another study,where the conversion rate was reducedwith further addition of sulfuricacid after themaximum conversion ratewas achieved at a certain sulfu-ric acid amount [14,27]. This may be due to the excess quantity of sulfu-ric acid consumes more KOH during neutralization, but is accounted asFFA.

3.2. Comparison of chemical usage

Fig. 4(a and b) summarizes the optimal chemical usages and initialFFA levels from a combination of literature [14–31], the NREL suggestedvalue used by biodiesel producers, and data from this study. Fig. 4a indi-cates that themethanol dosage is reasonably correlated to the FFA level,with an R2 of 0.77. It is also suggested that the NREL suggested dose issuitable for the oil with 15 to 25% FFA. For the oil with FFA higher

than 25%, the 20:1 methanol ratio is overdosed, which results in unnec-essary methanol input and more energy input for methanol recovery.For FFA less than 15%, the methanol ratio is underdosed and the esteri-fication reaction may not be complete in the desired time period. Asshown in this study, when using 20:1 methanol-to-FFA ratio (Fig. 1),none of the experiments could reach the target yield in the reactiontime window. The longer reaction time was necessary to complete thereaction, which required extra energy input and slowed down the pro-duction. Unfortunately, the major underdosed range lies in the yellowgrease range, which is defined as the oil with FFA less than 15% by bio-diesel industry. Generally, the acid esterification could pretreat the oilwith FFA up to 20%, so the biodiesel industry strongly prefers yellowgrease to brown grease. For oils with less than 15% FFA, an optimizationstudy is necessary to not only maintain high reaction yield and goodfinal product quality, but also lower the chemical cost and energy con-sumption. Fig. 4b shows the optimal acid usages and initial FFA levelsfrom various studies [13,14,18–20,22–27,29–31]. Similar to the metha-nol usage, the relationship between optimal acid amount and the initialFFA percentage is not linear. The lower R2 for the regression suggeststhat the FFA to acid correlation is not as strong as that of FFA to metha-nol. In themiddle FFA level, roughly between 15 and 35%, the NREL sug-gested dose is close to the regression curve, while in the low FFA range(b15%), NREL dose is underdosed and in the high FFA range (N35%),NREL dose is overdosed. Again, this suggests the necessity of optimiza-tion tests for yellow grease.

Two biodiesel producers, Bluegrass Biodiesel (10 million gallonsper year production capacity) and Greenleaf Biofuels (20 gpm feed-stock flow rate), have used both the NREL dosing suggestion, and theoptimized dosage based on the regression model (Fig. 4a and b) withtheir feedstock. Results shown in Table 4 indicated that using opti-mized dosage improved product yield as compared to the NREL sug-gested value when FFA ranges were outside of the effective NRELrange of 15–35%.

3.3. Parametric analysis

The experimental results indicated that the 2-hour FFA conversionrate could be affected by the reaction temperature, methanol and sulfu-ric acid quantities. To evaluate the relative contribution of each factor, amultivariable linear model and a multivariate quadratic model were

Fig. 3. Effect of acid catalyst amount on the 2-hour FFA conversion rate.

a b

Fig. 4. a. Comparison of methanol usage between laboratory and industrial practices. b. Comparison of H2SO4 usage between laboratory and industrial practices.

110 M. Chai et al. / Fuel Processing Technology 125 (2014) 106–113

employed. Similar approach was applied in other environmental dataanalysis [13,14]. The parametric analysis results could be used as aguide for industrial production to predict the reaction yields.

The linear model is expressed as the following Eq. (2).

y ¼ β0 þ β1T þ β2Aþ β3M ð2Þ

Where:

y the FFA conversion rate after two hours;βi (i = 0 ~ 3) coefficient for each variableT temperature (°C)A catalyst amount weight percentage to FFAM methanol-to-FFA molar ratio

The non-linear model is expressed as the following second-orderpolynomial Eq. (3).

y ¼ β0 þ β1T þ β2Aþ β3M þ β11T2 þ β22A

2 þ β33M2 þ β12T � A

þ β13T �M þ β23A�M ð3Þ

βjk (j = 1 ~ 3,k = 1 ~ 3) correlation coefficients for each variable to bedetermined.

Other denotations are the same as in Eq. (2).The results of analysis of variance (ANOVA) are listed in Table 5. Both

linear and non-linear models have high F-values, at 35.41 and 32.21 re-spectively. The R2 of the linear model is 0.57, and the non-linear modelhas an R2 of 0.77, which indicates that the non-linear model fits the ex-perimental data better. The correlation coefficients between the 2-hourFFA conversion rate and each contributing parameter in the non-linearmodel are listed in Table 6. The four highest correlation coefficientsare T × M (Temperature ∗ Methanol-to-FFA molar ratio), T (Tempera-ture), Methanol, and T2 (Temperature2) at 0.70, 0.57, 0.55, and 0.54 re-spectively. This indicates that reaction temperature has the highest

impact on FFA conversion rate, followed by methanol-to-FFA molarratio. These results are consistent with experimental observations, aswell as other studies in Fig. 4a and b. The comparison of experimentaland predicted two hour conversion rates is shown in Fig. 5. Almost90% of the experimental conversion rates (76 out of 85) are within±15% of the predicted range.

3.4. Reaction kinetics

The acid-catalyzed esterification is a reversible reaction. However,given the significantly excessive methanol use as compared with FFA,the reverse reaction could be neglected [13]. Some studies havesuggested that the experimental data could fit into a first-order reac-tion [13]. Hence, the activation energy can be determined by Arrheniusequation, shown below:

k ¼ Ae−EaRT ð4Þ

where: k is the reaction rate constant (min−1) derived from pseudo-first order reaction. A is the pre-exponential factor (min−1) and Eadenotes the activation energy (J∙mol−1). R and T stand for ideal gasconstant (8.314 J∙mol−1∙K−1) and temperature (K), respectively.

Ea and A can be determined by plotting the “ln(k) vs T−1” graph, andthe results are listed in Table 7. The R2 for all calculations are between0.95 and 0.99, which indicate a good fit of experimental data with theequation. With H2SO4 as the catalyst, the activation energy rangedfrom 20.7 to 45.9 kJ/mol, which is within the range of the existing re-sults [13,15,33,34]. Usingmethanol-to-FFAmolar ratio of 40:1, the acti-vation energies decreased from 5 to 12.5% acid usage and then greatlyincreased at 15% acid usage.With thefixed acid usage of 10%, the activa-tion energy decreased with the increasingmethanol-to-FFAmolar ratiofrom 60:1 to 50:1 and then sharply increased. The low activation ener-gies at methanol-to-FFA molar ratio of 40:1 to 50:1 and 10–12.5%(weight relative to FFA) acid quantity were consistent with higher FFAconversion rate observed at these reaction conditions. Furthermore, ifwe were to use the NREL suggested methanol-to-FFA molar ratio of20:1 or 5% acid, the resultant Eawould be the largest, whichwould affectthe effectiveness of the pretreatment.

4. Conclusions

In this study, the differences of laboratory and industrial practices onthe FFA pretreatment reaction were evaluated, and the used vegetableoil with 5± 0.5% FFAwas pretreated via an acid-catalyzed esterificationto better understand the difference. The optimal condition of this usedcooking oil: 55–65 °C, themethanol-to-FFAmolar ratio of 40:1, and sul-furic acid (catalyst) usage of 10% (weight relative to FFA), fell out of thesuggested values used by the biodiesel industry. It is found that the sug-gested 20:1methanol-to-FFAmolar ratio workedwell within FFA rangeof 15–25% and the suggested 5% acid amount worked well within theFFA range of 15–35%. Outside these ranges, an optimization study isnecessary since the esterification reaction might not proceed withthese conditions. For yellow grease (FFA less than 15%), the suggested20:1 methanol-to-FFA molar ratio and 5% acid use may result inunderdose. The correlation between the initial FFA level and themethanol-to-FFA ratio was found stronger than that between the initial

Table 4Yield of esterfication comparison between the suggested and optimized dosages by two biodiesel producers.

Initial FFA content NREL dose Optimized dosage from Fig. 4 (tested in two biodiesel plants)

Methanol-to-FFA molar ratio Catalyst-to-FFA wt% Yield%

Methanol-to-FFA molar ratio Catalyst-to-FFA wt% Yield%

2.5% 19.8:1 5% 80.0 89.3 19.8 92.06% 86.7 48.9 12.4 95.810% 90.0 34.3 9.4 96.6

Table 5Statistical analyses of linear and non-linear regression models.

Linear model Non-linear model

Coefficient p-Value Coefficient p-Value

β0 0.105 0.214 −1.25 1.36E−5β1 0.0083 4.76E−10 0.058 7.5E−8β2 −0.9E−3 0.827 – –

β3 0.0073 1.76E−9 0.016 0.03β11 −0.4E−3 0.16E−3β22 −0.004 5.64E−5β33 −0.3E−3 5.97E−7β12 −0.5E−3 0.198β13 −0.2E−3 0.029β23 0.0027 0.1E−3R2 0.567 0.77Standard error 0.106 0.080F-value 35.41 32.21Significance F b0.001 b0.001

111M. Chai et al. / Fuel Processing Technology 125 (2014) 106–113

FFA and sulfuric acid dosages, indicating quantity of methanol ismore crucial to the esterification reaction. The activation energiesof the esterification reaction for the test oil ranged from 20.7 to45.9 kJ/mol. The lowest activation energy was obtained with the op-timized chemical dosage, while the highest activation energy wasobtained under the operation conditions used by the industry(NREL recommended dosages).

Disclaimer

The U.S. Environmental Protection Agency, through its Office of Re-search and Development, funded and managed, or partially fundedand collaborated in, the research described herein, it has been subjectedto theAgency administrative review andhas been approved for externalpublication. Any opinions expressed in this paper are those of the au-thor(s) and do not necessarily reflect the views of the Agency, therefore,no official endorsement should be inferred. Anymention of trade namesor commercial products does not constitute endorsement or recom-mendation for use.

Acknowledgments

The authors acknowledge the financial supports from the RET(Research Experiences for Teachers) program of the National ScienceFoundation (EEC 0808696, EEC 1004623) and the National Risk Man-agement Research Laboratory, U.S. Environmental Protection Agency.The Cincinnati Zoo and Botanical Garden is acknowledged for providingthe used cooking oil.

References

[1] M.S. Graboski, R.L. McCormick, Combustion of fat and vegetable oil derived fuels indiesel engines, Progress in Energy and Combustion Science 24 (1998) 125–164.

[2] J. Sheehan, V. Camobreco, J. Duffield, M. Graboski, H. Shapouri, An overview of bio-diesel and petroleum diesel life cycles, Report NREL/TP-580-24772, NREL, Golden,CO, 1998.

[3] G.K. Casman, J.A. Liska, Food and fuel for all: realistic or foolish? Biofuels,Bioproducts and Biorefining 1 (2007) 18–23.

[4] R. Dominguez-faus, S.E. Powers, J.G. Burken, P.J. Alvarez, The water footprint ofbiofuels: a drink or drive issue? Environmental Science and Technology 43 (2009)3005–3010.

[5] M.J. Haas, T.A. Foglia, Alternative feedstocks and technology for biodiesel produc-tion, in: G. Knothe, J. Van Gerpen (Eds.), The Biodiesel Handbook, AOCS Press, Ur-bana, IL, 2005, pp. 42–61.

[6] M. Canakci, The potential of restaurant waste lipids as biodiesel feedstocks,Bioresource Technology 98 (2007) 183–190.

[7] R. Agnew, M. Chai, M. Lu, N. Dendramis, Making biodiesel from recycled cooking oilgenerated in campus dining facilities, Sustainability 2 (2009) 303–307.

[8] W. Diaz-Felix,M.R. Riley,W. Zimmt,M. Kazz, Pretreatment of yellowgrease for efficientproduction of fatty acid methyl esters, Biomass and Bioenergy 33 (2009) 558–563.

[9] M.G. Kulkarni, A.K. Dalai, Waste cooking oil — an economical source for biodiesel: areview, Industrial and Engineering Chemistry Research 45 (2006) 2901–2913.

[10] F. Ma, L.D. Clements, M.A. Hanna, The effects of catalyst free fatty acids and water ontransesterification of beef tallow, Transactions of ASAE 41 (1998) 1261–1264.

[11] J. Van Gerpen, B. Shanks, R. Pruszko, D. Clements, G. Knothe, Biodiesel productiontechnology, NREL/SR-510-36244, NREL, Golden, CO, 2004.

[12] M. Canakci, J. Van Gerpen, Biodiesel production via acid catalysis, Transactions ofASAE 42 (1999) 1203–1210.

[13] M. Berrios, J. Siles, M.A. Martin, A. Martin, A kinetic study of the esterification of freefatty acids (FFA) in sunflower oil, Fuel 86 (2007) 2383–2388.

[14] A.S. Ramadhas, S. Jayaraj, C. Muraleedharan, Biodiesel production from high FFArubber seed oil, Fuel 84 (2005) 335–340.

[15] K.V. Thiruvengadaravi, J. Nandagopal, V. Sathya Selva Bala, S. Dinesh Kirupha, P.Vijayalakshmi, S. Sivanesan, Kinetic study of the esterification of free fatty acids innon-edible Pongamia pinnata oil using acid catalyst, Indian Journal of Science andTechnology 2 (2009) 20–24.

[16] S.V. Ghadge, H. Raheman, Biodiesel production from Mahua (Madhuca indica) oilhaving high free fatty acids, Biomass and Bioenergy 28 (2005) 601–605.

[17] H.J. Berchmans, S. Hirata, Biodiesel production from crude Jatropha curcas L. seed oilwith a high content of free fatty acids, Bioresource Technology 99 (2008) 1716–1721.

[18] A.K. Tiwari, A. Kumar, H. Raheman, Biodiesel production from jatropha oil (Jatrophacurcas) with high free fatty acids: an optimized process, Biomass and Bioenergy 31(2007) 569–575.

[19] S. Chongkhong, C. Tongurai, P. Chetpattananondh, C. Bunyakan, Biodiesel produc-tion by esterification of palm fatty acid distillate, Biomass and Bioenergy 31(2007) 563–568.

[20] L. Lin, D. Ying, S. Chaitep, S. Vittayapadung, Biodiesel production from crude ricebran oil and properties as fuel, Applied Energy 86 (2009) 681–688.

[21] K. Suppalakpanya, S.B. Ratanawilai, C. Tongurai, Production of ethyl ester from crudepalm oil by two-step reaction with a microwave system, Fuel 89 (2010) 2140–2144.

[22] V.B. Veljkovic, S.H. Lakicevic, O.S. Stamenkovic, Z.B. Todorovic, M.L. Lazic, Biodieselproduction from tobacco (Nicotiana tabacum L.) seed oil with a high content offree fatty acids, Fuel 85 (2006) 2671–2675.

[23] M.A. Khan, S. Yusup, M.M. Ahmad, Acid esterification of a high free fatty acid crudepalm oil and crude rubber oil blend: optimization and parametric analysis, Biomassand Bioenergy 34 (2010) 1751–1756.

[24] H.A. Farage, A. El-Maghraby, N.A. Taha, Optimization of factors affecting esterifica-tion of mixed oil with high percentage of free fatty acid, Fuel Processing Technology92 (2011) 507–510.

[25] A. Hayyan, M.Z. Alam, M.E.S. Mirghani, N.A. Kabbashi, N.I.N.M. Hakimi, Y.M. Siran, S.Tahiruddin, Reduction of high content of free fatty acid in sludge palm oil via acidcatalyst for biodiesel production, Fuel Processing Technology 92 (2011) 920–924.

[26] S. Jansri, S.B. Ratanawilai, M.L. Allen, G. Prateepchaikul, Kinetics of methyl ester pro-duction from mixed crude palm oil by using acid–alkali catalyst, Fuel ProcessingTechnology 92 (2011) 1543–1548.

Table 6Correlation coefficients between 2-hour FFA conversion rate and factors.

T A M T2 A2 M2 T × A T × M A × M

2-hour FFA conversion rate 0.57 0.0073 0.55 0.54 −0.011 0.49 0.31 0.70 0.40

0%

20%

40%

60%

80%

100%

120%

0% 20% 40% 60% 80% 100% 120%

Experimental 2-hour FFA conversion rate

Pre

dict

ed 2

-hou

r F

FA

con

vers

ion

rate

from

non-

linea

r m

odel

Data Points

Theoretical Line

Fig. 5. Comparison of predicted and experimental the 2-hour FFA conversion rate.

Table 7Energies of activation and frequency factors of acid-catalyzed esterification of FFA.

Methanol-to-FFA molarratio

Sulfuric acid % wt toFFA

Ea (J/mol)

A R2

40 15% 38798.94 19747. 85 0.9940 12.5% 20746.76 43.25 0.9940 10% 24440.67 207.41 0.9640 7.5% 34370.91 6666.83 0.9640 5% 42007.32 97831.32 0.9560 10% 29235.35 1241.90 0.9650 10% 23132.04 128.32 0.9740 10% 24440.67 207.41 0.9630 10% 43106.43 150843.8 0.9520 10% 45936.51 179512.5 0.96

112 M. Chai et al. / Fuel Processing Technology 125 (2014) 106–113

[27] M. Charoenchaitrakool, J. Thienmethangkoon, Statistical optimization for biodieselproduction from waste frying oil through two-step catalyzed process, Fuel Process-ing Technology 92 (2011) 112–118.

[28] H.V. Kamath, I. Regupathi,M.B. Saidutta,Optimizationof two stepkaranja biodiesel syn-thesis under microwave irradiation, Fuel Processing Technology 92 (2011) 100–105.

[29] A. Hayyan, M.Z. Alam, M.E.S. Mirghani, N.A. Kabbashi, N.I.N.M. Hakimi, Y.M. Siran, S.Tahiruddin, Sludge palm oil as a renewable rawmaterial for biodiesel production bytwo-step process, Bioresource Technology 101 (2010) 7804–7811.

[30] Y. Wang, S. Ou, P. Liu, F. Xue, S. Tang, Comparison of two different processes to syn-thesize biodiesel by waste cooking oil, Journal of Molecular Catalysis A: Chemical252 (2006) 107–112.

[31] P.D. Patil, V.G. Gude, H.K. Reddy, T. Muppaneni, S. Deng, Biodiesel production fromwaste cooking oil using sulfuric acid and microwave irradiation processes, Journalof Environmental Protection 3 (2012) 107–113.

[32] American Oil Chemist's Society, AOCSmethods for biodiesel feedstock quality, Tech-nical Services AOCS (CD-ROM), AOCS press, Urbana, IL, 2007.

[33] E. Sendzikiene, V. Makareviciene, P. Janulis, S. Kitrys, Kinetics of free fatty acids es-terification with methanol in the production of biodiesel fuel, European Journal ofLipid Science and Technology 106 (2004) 831–836.

[34] Satriana,M.D. Supardan, Kinetic study of esterification of free fatty acid in low-gradecrude palm oil using sulfuric acid, ASEAN Journal of Chemical Engineering 8 (2008)1–8

113M. Chai et al. / Fuel Processing Technology 125 (2014) 106–113